X-ray lens figure errors retrieved by deep learning from several beam intensity images

Area of Science:

• Optics and Photonics
• Materials Science
• Computational Physics

Background:

• The phase problem is a significant challenge in X-ray optics, particularly for focusing synchrotron beams.
• Accurate characterization of lens aberrations is crucial for high-resolution imaging and experiments.

Purpose of the Study:

• To address the phase problem in X-ray lens systems.
• To demonstrate the feasibility of retrieving lens surface errors using only intensity measurements.
• To develop a method for routine aberration measurement in X-ray optical systems.

Main Methods:

• Simulations of X-ray beam propagation through lenses with random surface errors.
• Utilizing intensity images of the propagated beam at multiple distances.
• Training a neural network to predict lens error profiles from intensity data.

Main Results:

• Successfully demonstrated the retrieval of lens surface errors from intensity images alone.
• A trained neural network accurately predicts the lens error profile, accounting for all aberrations.
• The method shows feasibility for routine measurement of X-ray lens aberrations.

Conclusions:

• Intensity-based measurement is a viable solution to the phase problem in X-ray lens systems.
• Neural networks offer an efficient tool for characterizing optical aberrations.
• This technique can be applied to X-ray lenses and other optical systems for aberration measurement.